Ocean Modeling - EAS 880338 CHAPTER2. THECORIOLISFORCE y x X r Y i j J I Ωt Figure 2-1 Fixed (X, Y...
Transcript of Ocean Modeling - EAS 880338 CHAPTER2. THECORIOLISFORCE y x X r Y i j J I Ωt Figure 2-1 Fixed (X, Y...
Ocean Modeling - EAS 8803
We model the ocean on a rotating planet
Rotation effects are considered through the Coriolis and Centrifugal Force
The Coriolis Force arises because our reference frame (the Earth) is rotating
The Coriolis Force is the source of many interesting geophysical processes
Chapter 2
A rotating framework - The coordinates38 CHAPTER 2. THE CORIOLIS FORCE
!
!
"
#
$
$
y
x
X
r
Y i
j J
I
$"
#
!ΩtFigure 2-1 Fixed (X, Y ) and rotating(x, y) frameworks of reference.
plane are related by
x = + X cosΩt + Y sin Ωt (2.3a)
y = − X sin Ωt + Y cosΩt. (2.3b)
The first time derivative of the preceding expressions yields
dx
dt= +
dX
dtcosΩt +
dY
dtsin Ωt
+Ωy︷ ︸︸ ︷− ΩX sinΩt + ΩY cosΩt (2.4a)
dy
dt= −
dX
dtsin Ωt +
dY
dtcosΩt − ΩX cosΩt − ΩY sin Ωt︸ ︷︷ ︸
−Ωx
. (2.4b)
The quantities dx/dt and dy/dt give the rates of change of the coordinates relative to themoving frame as time evolves. They are thus the components of the relative velocity:
u =dx
dti +
dy
dtj = ui + vj. (2.5)
Similarly, dX/dt and dY/dt give the rates of change of the absolute coordinates and formthe absolute velocity:
U =dX
dtI +
dY
dtJ.
Writing the absolute velocity in terms of the rotating unit vectors, we obtain [using (2.2)]
U =
(dX
dtcosΩt +
dY
dtsin Ωt
)i +
(−
dX
dtsin Ωt +
dY
dtcosΩt
)j
= U i + V j. (2.6)
38 CHAPTER 2. THE CORIOLIS FORCE
!
!
"
#
$
$
y
x
X
r
Y i
j J
I
$"
#
!ΩtFigure 2-1 Fixed (X, Y ) and rotating(x, y) frameworks of reference.
plane are related by
x = + X cosΩt + Y sin Ωt (2.3a)
y = − X sin Ωt + Y cosΩt. (2.3b)
The first time derivative of the preceding expressions yields
dx
dt= +
dX
dtcosΩt +
dY
dtsin Ωt
+Ωy︷ ︸︸ ︷− ΩX sinΩt + ΩY cosΩt (2.4a)
dy
dt= −
dX
dtsin Ωt +
dY
dtcosΩt − ΩX cosΩt − ΩY sin Ωt︸ ︷︷ ︸
−Ωx
. (2.4b)
The quantities dx/dt and dy/dt give the rates of change of the coordinates relative to themoving frame as time evolves. They are thus the components of the relative velocity:
u =dx
dti +
dy
dtj = ui + vj. (2.5)
Similarly, dX/dt and dY/dt give the rates of change of the absolute coordinates and formthe absolute velocity:
U =dX
dtI +
dY
dtJ.
Writing the absolute velocity in terms of the rotating unit vectors, we obtain [using (2.2)]
U =
(dX
dtcosΩt +
dY
dtsin Ωt
)i +
(−
dX
dtsin Ωt +
dY
dtcosΩt
)j
= U i + V j. (2.6)
Fixed reference
Rotating reference
Angular Velocity
A rotating framework - The velocity (1st derivative)
38 CHAPTER 2. THE CORIOLIS FORCE
!
!
"
#
$
$
y
x
X
r
Y i
j J
I
$"
#
!ΩtFigure 2-1 Fixed (X, Y ) and rotating(x, y) frameworks of reference.
plane are related by
x = + X cosΩt + Y sin Ωt (2.3a)
y = − X sin Ωt + Y cosΩt. (2.3b)
The first time derivative of the preceding expressions yields
dx
dt= +
dX
dtcosΩt +
dY
dtsin Ωt
+Ωy︷ ︸︸ ︷− ΩX sinΩt + ΩY cosΩt (2.4a)
dy
dt= −
dX
dtsin Ωt +
dY
dtcosΩt − ΩX cosΩt − ΩY sin Ωt︸ ︷︷ ︸
−Ωx
. (2.4b)
The quantities dx/dt and dy/dt give the rates of change of the coordinates relative to themoving frame as time evolves. They are thus the components of the relative velocity:
u =dx
dti +
dy
dtj = ui + vj. (2.5)
Similarly, dX/dt and dY/dt give the rates of change of the absolute coordinates and formthe absolute velocity:
U =dX
dtI +
dY
dtJ.
Writing the absolute velocity in terms of the rotating unit vectors, we obtain [using (2.2)]
U =
(dX
dtcosΩt +
dY
dtsin Ωt
)i +
(−
dX
dtsin Ωt +
dY
dtcosΩt
)j
= U i + V j. (2.6)
Absolute Velocity
Relative Velocitychange of the coordinate relative to the moving frame
38 CHAPTER 2. THE CORIOLIS FORCE
!
!
"
#
$
$
y
x
X
r
Y i
j J
I
$"
#
!ΩtFigure 2-1 Fixed (X, Y ) and rotating(x, y) frameworks of reference.
plane are related by
x = + X cosΩt + Y sin Ωt (2.3a)
y = − X sin Ωt + Y cosΩt. (2.3b)
The first time derivative of the preceding expressions yields
dx
dt= +
dX
dtcosΩt +
dY
dtsin Ωt
+Ωy︷ ︸︸ ︷− ΩX sinΩt + ΩY cosΩt (2.4a)
dy
dt= −
dX
dtsin Ωt +
dY
dtcosΩt − ΩX cosΩt − ΩY sin Ωt︸ ︷︷ ︸
−Ωx
. (2.4b)
The quantities dx/dt and dy/dt give the rates of change of the coordinates relative to themoving frame as time evolves. They are thus the components of the relative velocity:
u =dx
dti +
dy
dtj = ui + vj. (2.5)
Similarly, dX/dt and dY/dt give the rates of change of the absolute coordinates and formthe absolute velocity:
U =dX
dtI +
dY
dtJ.
Writing the absolute velocity in terms of the rotating unit vectors, we obtain [using (2.2)]
U =
(dX
dtcosΩt +
dY
dtsin Ωt
)i +
(−
dX
dtsin Ωt +
dY
dtcosΩt
)j
= U i + V j. (2.6)
38 CHAPTER 2. THE CORIOLIS FORCE
!
!
"
#
$
$
y
x
X
r
Y i
j J
I
$"
#
!ΩtFigure 2-1 Fixed (X, Y ) and rotating(x, y) frameworks of reference.
plane are related by
x = + X cosΩt + Y sin Ωt (2.3a)
y = − X sin Ωt + Y cosΩt. (2.3b)
The first time derivative of the preceding expressions yields
dx
dt= +
dX
dtcosΩt +
dY
dtsin Ωt
+Ωy︷ ︸︸ ︷− ΩX sinΩt + ΩY cosΩt (2.4a)
dy
dt= −
dX
dtsin Ωt +
dY
dtcosΩt − ΩX cosΩt − ΩY sin Ωt︸ ︷︷ ︸
−Ωx
. (2.4b)
The quantities dx/dt and dy/dt give the rates of change of the coordinates relative to themoving frame as time evolves. They are thus the components of the relative velocity:
u =dx
dti +
dy
dtj = ui + vj. (2.5)
Similarly, dX/dt and dY/dt give the rates of change of the absolute coordinates and formthe absolute velocity:
U =dX
dtI +
dY
dtJ.
Writing the absolute velocity in terms of the rotating unit vectors, we obtain [using (2.2)]
U =
(dX
dtcosΩt +
dY
dtsin Ωt
)i +
(−
dX
dtsin Ωt +
dY
dtcosΩt
)j
= U i + V j. (2.6)
38 CHAPTER 2. THE CORIOLIS FORCE
!
!
"
#
$
$
y
x
X
r
Y i
j J
I
$"
#
!ΩtFigure 2-1 Fixed (X, Y ) and rotating(x, y) frameworks of reference.
plane are related by
x = + X cosΩt + Y sin Ωt (2.3a)
y = − X sin Ωt + Y cosΩt. (2.3b)
The first time derivative of the preceding expressions yields
dx
dt= +
dX
dtcosΩt +
dY
dtsin Ωt
+Ωy︷ ︸︸ ︷− ΩX sinΩt + ΩY cosΩt (2.4a)
dy
dt= −
dX
dtsin Ωt +
dY
dtcosΩt − ΩX cosΩt − ΩY sin Ωt︸ ︷︷ ︸
−Ωx
. (2.4b)
The quantities dx/dt and dy/dt give the rates of change of the coordinates relative to themoving frame as time evolves. They are thus the components of the relative velocity:
u =dx
dti +
dy
dtj = ui + vj. (2.5)
Similarly, dX/dt and dY/dt give the rates of change of the absolute coordinates and formthe absolute velocity:
U =dX
dtI +
dY
dtJ.
Writing the absolute velocity in terms of the rotating unit vectors, we obtain [using (2.2)]
U =
(dX
dtcosΩt +
dY
dtsin Ωt
)i +
(−
dX
dtsin Ωt +
dY
dtcosΩt
)j
= U i + V j. (2.6)
U
V
u
v
2.1. ROTATING FRAMEWORK 39
Thus, dX/dt and dY/dt are the components of the absolute velocityU in the inertial frame,
whereas U and V are the components of the same vector in the rotating frame. Use of (2.4)
and (2.3) in the preceding expression yields the following relations between absolute and
relative velocities:
U = u − Ωy, V = v + Ωx. (2.7)
These equalities simply state that the absolute velocity is the relative velocity plus the en-
training velocity due to the rotation of the reference framework.
A second derivative with respect to time provides in a similar manner:
d2x
dt2=
(d2X
dt2cosΩt +
d2Y
dt2sin Ωt
)+ 2Ω
(−
dX
dtsin Ωt +
dY
dtcosΩt
)
︸ ︷︷ ︸V
− Ω2 (X cosΩt + Y sin Ωt)︸ ︷︷ ︸x
(2.8a)
d2y
dt2=
(−
d2X
dt2sin Ωt +
d2Y
dt2cosΩt
)− 2Ω
(dX
dtcosΩt +
dY
dtsin Ωt
)
︸ ︷︷ ︸U
− Ω2 (− X sin Ωt + Y cosΩt)︸ ︷︷ ︸y
. (2.8b)
Expressed in terms of the relative and absolute accelerations
a =d2x
dt2i +
d2y
dt2j =
du
dti +
dv
dtj = ai + bj
A =d2X
dt2I +
d2Y
dt2J
=
(d2X
dt2cosΩt +
d2Y
dt2sin Ωt
)i +
(d2Y
dt2cosΩt −
d2X
dt2sin Ωt
)j = Ai + Bj,
expressions (2.8) condense to
a = A + 2ΩV − Ω2x, b = B − 2ΩU − Ω2y.
In analogy with the absolute velocity vector, d2X/dt2 and d2Y/dt2 are the components ofthe absolute accelerationA in the inertial frame, whereasA and B are the components of the
same vector in the rotating frame. The absolute acceleration components, necessary later to
formulate Newton’s law, are obtained by solving for A and B and using (2.7):
A = a − 2Ωv − Ω2x, B = b + 2Ωu − Ω2y. (2.9)
We now see that the difference between absolute and relative acceleration consists of two
contributions. The first, proportional to Ω and to the relative velocity, is called the Coriolis
acceleration; the other, proportional to Ω2 and to the coordinates, is called the centrifugal
acceleration. When placed on the other side of the equality in Newton’s law, these terms
Relation between absolute and relative velocity
Ωy
−Ωx
entraining velocity due to rotation
relative velocity +absolute
velocity =
A rotating framework - The velocity (1st derivative)
A rotating framework - The acceleration (2nd derivative)
2.1. ROTATING FRAMEWORK 39
Thus, dX/dt and dY/dt are the components of the absolute velocityU in the inertial frame,
whereas U and V are the components of the same vector in the rotating frame. Use of (2.4)
and (2.3) in the preceding expression yields the following relations between absolute and
relative velocities:
U = u − Ωy, V = v + Ωx. (2.7)
These equalities simply state that the absolute velocity is the relative velocity plus the en-
training velocity due to the rotation of the reference framework.
A second derivative with respect to time provides in a similar manner:
d2x
dt2=
(d2X
dt2cosΩt +
d2Y
dt2sin Ωt
)+ 2Ω
(−
dX
dtsin Ωt +
dY
dtcosΩt
)
︸ ︷︷ ︸V
− Ω2 (X cosΩt + Y sin Ωt)︸ ︷︷ ︸x
(2.8a)
d2y
dt2=
(−
d2X
dt2sin Ωt +
d2Y
dt2cosΩt
)− 2Ω
(dX
dtcosΩt +
dY
dtsin Ωt
)
︸ ︷︷ ︸U
− Ω2 (− X sin Ωt + Y cosΩt)︸ ︷︷ ︸y
. (2.8b)
Expressed in terms of the relative and absolute accelerations
a =d2x
dt2i +
d2y
dt2j =
du
dti +
dv
dtj = ai + bj
A =d2X
dt2I +
d2Y
dt2J
=
(d2X
dt2cosΩt +
d2Y
dt2sin Ωt
)i +
(d2Y
dt2cosΩt −
d2X
dt2sin Ωt
)j = Ai + Bj,
expressions (2.8) condense to
a = A + 2ΩV − Ω2x, b = B − 2ΩU − Ω2y.
In analogy with the absolute velocity vector, d2X/dt2 and d2Y/dt2 are the components ofthe absolute accelerationA in the inertial frame, whereasA and B are the components of the
same vector in the rotating frame. The absolute acceleration components, necessary later to
formulate Newton’s law, are obtained by solving for A and B and using (2.7):
A = a − 2Ωv − Ω2x, B = b + 2Ωu − Ω2y. (2.9)
We now see that the difference between absolute and relative acceleration consists of two
contributions. The first, proportional to Ω and to the relative velocity, is called the Coriolis
acceleration; the other, proportional to Ω2 and to the coordinates, is called the centrifugal
acceleration. When placed on the other side of the equality in Newton’s law, these terms
2.1. ROTATING FRAMEWORK 39
Thus, dX/dt and dY/dt are the components of the absolute velocityU in the inertial frame,
whereas U and V are the components of the same vector in the rotating frame. Use of (2.4)
and (2.3) in the preceding expression yields the following relations between absolute and
relative velocities:
U = u − Ωy, V = v + Ωx. (2.7)
These equalities simply state that the absolute velocity is the relative velocity plus the en-
training velocity due to the rotation of the reference framework.
A second derivative with respect to time provides in a similar manner:
d2x
dt2=
(d2X
dt2cosΩt +
d2Y
dt2sin Ωt
)+ 2Ω
(−
dX
dtsin Ωt +
dY
dtcosΩt
)
︸ ︷︷ ︸V
− Ω2 (X cosΩt + Y sin Ωt)︸ ︷︷ ︸x
(2.8a)
d2y
dt2=
(−
d2X
dt2sin Ωt +
d2Y
dt2cosΩt
)− 2Ω
(dX
dtcosΩt +
dY
dtsin Ωt
)
︸ ︷︷ ︸U
− Ω2 (− X sin Ωt + Y cosΩt)︸ ︷︷ ︸y
. (2.8b)
Expressed in terms of the relative and absolute accelerations
a =d2x
dt2i +
d2y
dt2j =
du
dti +
dv
dtj = ai + bj
A =d2X
dt2I +
d2Y
dt2J
=
(d2X
dt2cosΩt +
d2Y
dt2sin Ωt
)i +
(d2Y
dt2cosΩt −
d2X
dt2sin Ωt
)j = Ai + Bj,
expressions (2.8) condense to
a = A + 2ΩV − Ω2x, b = B − 2ΩU − Ω2y.
In analogy with the absolute velocity vector, d2X/dt2 and d2Y/dt2 are the components ofthe absolute accelerationA in the inertial frame, whereasA and B are the components of the
same vector in the rotating frame. The absolute acceleration components, necessary later to
formulate Newton’s law, are obtained by solving for A and B and using (2.7):
A = a − 2Ωv − Ω2x, B = b + 2Ωu − Ω2y. (2.9)
We now see that the difference between absolute and relative acceleration consists of two
contributions. The first, proportional to Ω and to the relative velocity, is called the Coriolis
acceleration; the other, proportional to Ω2 and to the coordinates, is called the centrifugal
acceleration. When placed on the other side of the equality in Newton’s law, these terms
dudt
dUdt 2ΩV
− Ω2xdvdt
dVdt −2ΩU
− Ω2y
A rotating framework - The acceleration (2nd derivative)
2.1. ROTATING FRAMEWORK 39
Thus, dX/dt and dY/dt are the components of the absolute velocityU in the inertial frame,
whereas U and V are the components of the same vector in the rotating frame. Use of (2.4)
and (2.3) in the preceding expression yields the following relations between absolute and
relative velocities:
U = u − Ωy, V = v + Ωx. (2.7)
These equalities simply state that the absolute velocity is the relative velocity plus the en-
training velocity due to the rotation of the reference framework.
A second derivative with respect to time provides in a similar manner:
d2x
dt2=
(d2X
dt2cosΩt +
d2Y
dt2sin Ωt
)+ 2Ω
(−
dX
dtsin Ωt +
dY
dtcosΩt
)
︸ ︷︷ ︸V
− Ω2 (X cosΩt + Y sin Ωt)︸ ︷︷ ︸x
(2.8a)
d2y
dt2=
(−
d2X
dt2sin Ωt +
d2Y
dt2cosΩt
)− 2Ω
(dX
dtcosΩt +
dY
dtsin Ωt
)
︸ ︷︷ ︸U
− Ω2 (− X sin Ωt + Y cosΩt)︸ ︷︷ ︸y
. (2.8b)
Expressed in terms of the relative and absolute accelerations
a =d2x
dt2i +
d2y
dt2j =
du
dti +
dv
dtj = ai + bj
A =d2X
dt2I +
d2Y
dt2J
=
(d2X
dt2cosΩt +
d2Y
dt2sin Ωt
)i +
(d2Y
dt2cosΩt −
d2X
dt2sin Ωt
)j = Ai + Bj,
expressions (2.8) condense to
a = A + 2ΩV − Ω2x, b = B − 2ΩU − Ω2y.
In analogy with the absolute velocity vector, d2X/dt2 and d2Y/dt2 are the components ofthe absolute accelerationA in the inertial frame, whereasA and B are the components of the
same vector in the rotating frame. The absolute acceleration components, necessary later to
formulate Newton’s law, are obtained by solving for A and B and using (2.7):
A = a − 2Ωv − Ω2x, B = b + 2Ωu − Ω2y. (2.9)
We now see that the difference between absolute and relative acceleration consists of two
contributions. The first, proportional to Ω and to the relative velocity, is called the Coriolis
acceleration; the other, proportional to Ω2 and to the coordinates, is called the centrifugal
acceleration. When placed on the other side of the equality in Newton’s law, these terms
dudt
dUdt 2ΩV
− Ω2x2.1. ROTATING FRAMEWORK 39
Thus, dX/dt and dY/dt are the components of the absolute velocityU in the inertial frame,
whereas U and V are the components of the same vector in the rotating frame. Use of (2.4)
and (2.3) in the preceding expression yields the following relations between absolute and
relative velocities:
U = u − Ωy, V = v + Ωx. (2.7)
These equalities simply state that the absolute velocity is the relative velocity plus the en-
training velocity due to the rotation of the reference framework.
A second derivative with respect to time provides in a similar manner:
d2x
dt2=
(d2X
dt2cosΩt +
d2Y
dt2sin Ωt
)+ 2Ω
(−
dX
dtsin Ωt +
dY
dtcosΩt
)
︸ ︷︷ ︸V
− Ω2 (X cosΩt + Y sin Ωt)︸ ︷︷ ︸x
(2.8a)
d2y
dt2=
(−
d2X
dt2sin Ωt +
d2Y
dt2cosΩt
)− 2Ω
(dX
dtcosΩt +
dY
dtsin Ωt
)
︸ ︷︷ ︸U
− Ω2 (− X sin Ωt + Y cosΩt)︸ ︷︷ ︸y
. (2.8b)
Expressed in terms of the relative and absolute accelerations
a =d2x
dt2i +
d2y
dt2j =
du
dti +
dv
dtj = ai + bj
A =d2X
dt2I +
d2Y
dt2J
=
(d2X
dt2cosΩt +
d2Y
dt2sin Ωt
)i +
(d2Y
dt2cosΩt −
d2X
dt2sin Ωt
)j = Ai + Bj,
expressions (2.8) condense to
a = A + 2ΩV − Ω2x, b = B − 2ΩU − Ω2y.
In analogy with the absolute velocity vector, d2X/dt2 and d2Y/dt2 are the components ofthe absolute accelerationA in the inertial frame, whereasA and B are the components of the
same vector in the rotating frame. The absolute acceleration components, necessary later to
formulate Newton’s law, are obtained by solving for A and B and using (2.7):
A = a − 2Ωv − Ω2x, B = b + 2Ωu − Ω2y. (2.9)
We now see that the difference between absolute and relative acceleration consists of two
contributions. The first, proportional to Ω and to the relative velocity, is called the Coriolis
acceleration; the other, proportional to Ω2 and to the coordinates, is called the centrifugal
acceleration. When placed on the other side of the equality in Newton’s law, these terms
use this equality:
Relation between absolute and relative velocity
A rotating framework - The acceleration (2nd derivative)
Coriolisacceleration
relative acceleration +absolute
acceleration = Centrifugalacceleration+
2.1. ROTATING FRAMEWORK 39
Thus, dX/dt and dY/dt are the components of the absolute velocityU in the inertial frame,
whereas U and V are the components of the same vector in the rotating frame. Use of (2.4)
and (2.3) in the preceding expression yields the following relations between absolute and
relative velocities:
U = u − Ωy, V = v + Ωx. (2.7)
These equalities simply state that the absolute velocity is the relative velocity plus the en-
training velocity due to the rotation of the reference framework.
A second derivative with respect to time provides in a similar manner:
d2x
dt2=
(d2X
dt2cosΩt +
d2Y
dt2sin Ωt
)+ 2Ω
(−
dX
dtsin Ωt +
dY
dtcosΩt
)
︸ ︷︷ ︸V
− Ω2 (X cosΩt + Y sin Ωt)︸ ︷︷ ︸x
(2.8a)
d2y
dt2=
(−
d2X
dt2sin Ωt +
d2Y
dt2cosΩt
)− 2Ω
(dX
dtcosΩt +
dY
dtsin Ωt
)
︸ ︷︷ ︸U
− Ω2 (− X sin Ωt + Y cosΩt)︸ ︷︷ ︸y
. (2.8b)
Expressed in terms of the relative and absolute accelerations
a =d2x
dt2i +
d2y
dt2j =
du
dti +
dv
dtj = ai + bj
A =d2X
dt2I +
d2Y
dt2J
=
(d2X
dt2cosΩt +
d2Y
dt2sin Ωt
)i +
(d2Y
dt2cosΩt −
d2X
dt2sin Ωt
)j = Ai + Bj,
expressions (2.8) condense to
a = A + 2ΩV − Ω2x, b = B − 2ΩU − Ω2y.
In analogy with the absolute velocity vector, d2X/dt2 and d2Y/dt2 are the components ofthe absolute accelerationA in the inertial frame, whereasA and B are the components of the
same vector in the rotating frame. The absolute acceleration components, necessary later to
formulate Newton’s law, are obtained by solving for A and B and using (2.7):
A = a − 2Ωv − Ω2x, B = b + 2Ωu − Ω2y. (2.9)
We now see that the difference between absolute and relative acceleration consists of two
contributions. The first, proportional to Ω and to the relative velocity, is called the Coriolis
acceleration; the other, proportional to Ω2 and to the coordinates, is called the centrifugal
acceleration. When placed on the other side of the equality in Newton’s law, these terms
use this equality:
dUdt
=dudt
− 2Ωv − Ω2x
dVdt
=dvdt
− 2Ωu − Ω2y
A rotating framework - The acceleration (2nd derivative)
40 CHAPTER 2. THE CORIOLIS FORCE
can be assimilated to forces (per unit mass). The centrifugal force acts as an outward pull,
whereas the Coriolis force depends on the direction and magnitude of the relative velocity.
Formally, the preceding results could have been derived in a vector form. Defining the
vector rotation
Ω = Ωk,
where k is the unit vector in the third dimension (which is common to both systems of refer-ence), we can write (2.7) and (2.9) as
U = u + Ω × r
A = a + 2Ω × u + Ω × (Ω × r), (2.10)
where× indicates the vectorial product. This implies that taking a time derivative of a vectorwith respect to the inertial framework is equivalent to applying the operator
d
dt+ Ω ×
in the rotating framework of reference.
A very detailed exposition of the Coriolis and centrifugal accelerations can be found in the
book by Stommel and Moore (1989). In addition, the reader will find a historical perspective
in Ripa (1994).
2.2 Unimportance of the centrifugal force
Unlike the Coriolis force, which is proportional to the velocity, the centrifugal force depends
solely on the rotation rate and the distance of the particle to the rotation axis. Even at rest
with respect to the rotating planet, particles experience an outward pull. Yet, on the earth as
on other celestial bodies, no matter goes flying out to space. How is that possible? Obviously,
gravity keeps everything together.
In the absence of rotation, gravitational forces keep the matter together to form a spherical
body (with the denser materials at the center and the lighter ones on the periphery). The
outward pull caused by the centrifugal force distorts this spherical equilibrium, and the planet
assumes a slightly flattened shape. The degree of flattening is precisely that necessary to keep
the planet in equilibrium for its rotation rate.
The situation is depicted on Figure 2-2. By its nature, the centrifugal force is directed
outward, perpendicular to the axis of rotation, whereas the gravitational force points toward
the planet’s center. The resulting force assumes an intermediate direction, and this direction
is precisely the direction of the local vertical. Indeed, under this condition a loose particle
would have no tendency of its own to fly away from the planet. In other words, every particle
at rest on the surface will remain at rest unless it is subjected to additional forces.
The flattening of the earth, as well as that of other celestial bodies in rotation, is important
to neutralize the centrifugal force. But, this is not to say that it greatly distorts the geometry.
On the earth, for example, the distortion is very slight, because gravity by far exceeds the
Vector Notation
Coriolis acceleration Centrifugal acceleration
r =xy⎡
⎣⎢
⎤
⎦⎥
dudt
de!ne vectors: u =uv⎡
⎣⎢
⎤
⎦⎥
ddt
+Ω ×time derivative in rotating frame:
Coriolis vs. Centrifugal Force
Centrifugal Force depends only on location
Coriolis Force is active only when things move
dUdt
=dudt
− 2Ωv − Ω2x
dVdt
=dvdt
− 2Ωu − Ω2y
Centrifugal Force is unimportant for motions
2.2. CENTRIFUGAL FORCE 41
....... ....... ......... ............ ............... ................. ................... ...................... ................................ .................................. ..................................... .......................................................... ............................................................ ............................................................ .......................................................... .......................................................................
...........................................................................................................................................
...................................... ...... ....... ........ ......... ......... .................. ......... ......... ........ ............. .....................................
Ω
S
N
Net
Centrifugal
Gravity
Localvertical
Figure 2-2 How the flattening of the
rotating earth (grossly exaggerated in
this drawing) causes the gravitational
and centrifugal forces to combine into
a net force aligned with the local verti-
cal, so that equilibrium is reached.
centrifugal force; the terrestrial equatorial radius is 6378 km, slightly greater than its polar
radius of 6357 km. The shape of the rotating oblate earth is treated in detail by Stommel and
Moore (1989) and by Ripa (1994).
For the sake of simplicity in all that follows, we will call the gravitational force the resul-
tant force, aligned with the vertical and equal to the sum of the true gravitational force and
the centrifugal force. Due to inhomogeneous distributions of rocks and magma on earth, the
true gravitational force is not directed towards the center of the earth. For the same reason as
the centrifugal force has rendered the earth surface oblate, this inhomogeneous true gravity
has deformed the earth surface until the total (apparent) gravitational force is perpendicular to
it. The surface so obtained is called a geoid and can be interpreted as the surface of an ocean
at rest (with a continuous extension on land). This virtual continuous surface is perpendic-
ular at every point to the direction of gravity (including the centrifugal force) and forms an
equipotential surface, meaning that a particle moving on that surface undergoes no change in
potential energy. The value of this potential energy per unit mass is called the geopotential,
and the geoid is thus a surface of constant geopotential. This surface will be the reference
surface from which land elevations, (dynamic) sea surface elevations and ocean depth will be
defined. For more on the geoid, the reader is referred to Robinson (2004), Chapter 11.
In a rotating laboratory tank, the situation is similar but not identical. The rotation causes
a displacement of the fluid toward the periphery. This proceeds until the resulting inward
pressure gradient prevents any further displacement. Equilibrium then requires that at any
point on the surface, the downward gravitational force and the outward centrifugal force
combine into a resultant force normal to the surface (Figure 2-3), so that the surface becomes
an equipotential surface. Although the surface curvature is crucial in neutralizing the cen-
trifugal force, the vertical displacements are rather small. In a tank rotating at the rate of one
revolution every two seconds (30 rpm) and 40 cm in diameter, the difference in fluid height
between the rim and the center is a modest 2 cm.
geoidan equipotential surface
Free Motion on a rotating frame
Coriolis Force is active only when things move
dUdt
=dudt
− 2Ωv − Ω2x
dVdt
=dvdt
− 2Ωu − Ω2y
42 CHAPTER 2. THE CORIOLIS FORCE
Ω
S
Net
Gravity
Centrifugal
........................................... ....... ........ ......... .......... .......... .......... .......... .......... .......... ......... .........................................................
Figure 2-3 Equilibrium surface of a
rotating fluid in an open container. The
surface slope is such that gravitational
and centrifugal forces combine into a
net force everywhere aligned with the
local normal to the surface.
2.3 Free motion on a rotating plane
The preceding argument allows us to combine the centrifugal force with the gravitational
force, but the Coriolis force remains. To have an idea of what this force can cause, let us
examine the motion of a free particle, that is, a particle not subject to any external force
other than gravity, the “horizontal” component of which is cancelled by the centrifugal force
associated with the rotating earth. On the plane attached to the North Pole of the rotating
earth no external horizontal forces is thus present.
If the particle is free of any force, its acceleration in the inertial frame is nil, by New-
ton’s law. According to (2.9), with the centrifugal-acceleration terms no longer present, the
equations governing the velocity components of the particle are
du
dt− 2Ωv = 0,
dv
dt+ 2Ωu = 0. (2.11)
The general solution to this system of linear equations is
u = V sin(ft + φ), v = V cos(ft + φ), (2.12)
where f = 2Ω, called the Coriolis parameter, has been introduced for convenience, and Vand φ are two arbitrary constants of integration. Without loss of generality, V can always be
chosen as nonnegative. (Do not confuse this constant V with the y–component of the absolutevelocity introduced in Section 2.1.) A first result is that the particle speed (u2+v2)1/2 remains
unchanged in time. It is equal to V , a constant determined by the initial conditions.Although the speed remains unchanged, the components u and v do depend on time, im-
plying a change in direction. To document this curving effect, it is most instructive to derive
the trajectory of the particle. The coordinates of the particle position change, by definition of
42 CHAPTER 2. THE CORIOLIS FORCE
Ω
S
Net
Gravity
Centrifugal
........................................... ....... ........ ......... .......... .......... .......... .......... .......... .......... ......... .........................................................
Figure 2-3 Equilibrium surface of a
rotating fluid in an open container. The
surface slope is such that gravitational
and centrifugal forces combine into a
net force everywhere aligned with the
local normal to the surface.
2.3 Free motion on a rotating plane
The preceding argument allows us to combine the centrifugal force with the gravitational
force, but the Coriolis force remains. To have an idea of what this force can cause, let us
examine the motion of a free particle, that is, a particle not subject to any external force
other than gravity, the “horizontal” component of which is cancelled by the centrifugal force
associated with the rotating earth. On the plane attached to the North Pole of the rotating
earth no external horizontal forces is thus present.
If the particle is free of any force, its acceleration in the inertial frame is nil, by New-
ton’s law. According to (2.9), with the centrifugal-acceleration terms no longer present, the
equations governing the velocity components of the particle are
du
dt− 2Ωv = 0,
dv
dt+ 2Ωu = 0. (2.11)
The general solution to this system of linear equations is
u = V sin(ft + φ), v = V cos(ft + φ), (2.12)
where f = 2Ω, called the Coriolis parameter, has been introduced for convenience, and Vand φ are two arbitrary constants of integration. Without loss of generality, V can always be
chosen as nonnegative. (Do not confuse this constant V with the y–component of the absolutevelocity introduced in Section 2.1.) A first result is that the particle speed (u2+v2)1/2 remains
unchanged in time. It is equal to V , a constant determined by the initial conditions.Although the speed remains unchanged, the components u and v do depend on time, im-
plying a change in direction. To document this curving effect, it is most instructive to derive
the trajectory of the particle. The coordinates of the particle position change, by definition of
Free Motion on a rotating frame
42 CHAPTER 2. THE CORIOLIS FORCE
Ω
S
Net
Gravity
Centrifugal
........................................... ....... ........ ......... .......... .......... .......... .......... .......... .......... ......... .........................................................
Figure 2-3 Equilibrium surface of a
rotating fluid in an open container. The
surface slope is such that gravitational
and centrifugal forces combine into a
net force everywhere aligned with the
local normal to the surface.
2.3 Free motion on a rotating plane
The preceding argument allows us to combine the centrifugal force with the gravitational
force, but the Coriolis force remains. To have an idea of what this force can cause, let us
examine the motion of a free particle, that is, a particle not subject to any external force
other than gravity, the “horizontal” component of which is cancelled by the centrifugal force
associated with the rotating earth. On the plane attached to the North Pole of the rotating
earth no external horizontal forces is thus present.
If the particle is free of any force, its acceleration in the inertial frame is nil, by New-
ton’s law. According to (2.9), with the centrifugal-acceleration terms no longer present, the
equations governing the velocity components of the particle are
du
dt− 2Ωv = 0,
dv
dt+ 2Ωu = 0. (2.11)
The general solution to this system of linear equations is
u = V sin(ft + φ), v = V cos(ft + φ), (2.12)
where f = 2Ω, called the Coriolis parameter, has been introduced for convenience, and Vand φ are two arbitrary constants of integration. Without loss of generality, V can always be
chosen as nonnegative. (Do not confuse this constant V with the y–component of the absolutevelocity introduced in Section 2.1.) A first result is that the particle speed (u2+v2)1/2 remains
unchanged in time. It is equal to V , a constant determined by the initial conditions.Although the speed remains unchanged, the components u and v do depend on time, im-
plying a change in direction. To document this curving effect, it is most instructive to derive
the trajectory of the particle. The coordinates of the particle position change, by definition of
NOTE: the speed does not change with time yet u and v do change with time!
changes in u and v imply change in direction.
Inertial Oscillations
Trajectory of inertial oscillations
2.3. FREE MOTION 43
the vector velocity, according to dx/dt = u and dy/dt = v, and a second time integrationprovides
x = x0 −V
fcos(ft + φ) (2.13a)
y = y0 +V
fsin(ft + φ), (2.13b)
where x0 and y0 are additional constants of integration to be determined from the initial
coordinates of the particle. From the last relations, it follows directly that
(x − x0)2 + (y − y0)
2 =
(V
f
)2
. (2.14)
This implies that the trajectory is a circle centered at (x0, y0) and of radius V/|f |. Thesituation is depicted on Figure 2-4.
R = Vf
.....................................................................................................................................
............................................
......................
!V
Ω = f2
(x0, y0)
Figure 2-4 Inertial oscillation of a free
particle on a rotating plane. The or-
bital period is exactly half of the am-
bient revolution period. This figure has
been drawn with a positive Coriolis pa-
rameter, f , representative of the North-ern Hemisphere. If f were negative (asin the Southern Hemisphere), the parti-
cle would veer to the left.
In the absence of rotation (f = 0), this radius is infinite, and the particle follows a straightpath, as we could have anticipated. But, in the presence of rotation (f "= 0), the particle
turns constantly. A quick examination of (2.13) reveals that the particle turns to the right
(clockwise) if f is positive or to the left (counterclockwise) if f is negative. In sum, the ruleis that the particle turns in the sense opposite to that of the ambient rotation.
At this point, we may wonder whether this particle rotation is none other than the negative
of the ambient rotation, in such a way as to keep the particle at rest in the absolute frame of
reference. But, there are at least two reasons why this is not so. The first is that the coordinates
of the center of the particle’s circular path are arbitrary and are therefore not required to
coincide with those of the axis of rotation. The second and most compelling reason is that
the two frequencies of rotation are simply not the same: the ambient rotating plane completes
one revolution in a time equal to Ta = 2π/Ω, whereas the particle covers a full circle in atime equal to Tp = 2π/f = π/Ω, called inertial period. Thus, the particle goes around itsorbit twice as the plane accomplishes a single revolution.
2.3. FREE MOTION 43
the vector velocity, according to dx/dt = u and dy/dt = v, and a second time integrationprovides
x = x0 −V
fcos(ft + φ) (2.13a)
y = y0 +V
fsin(ft + φ), (2.13b)
where x0 and y0 are additional constants of integration to be determined from the initial
coordinates of the particle. From the last relations, it follows directly that
(x − x0)2 + (y − y0)
2 =
(V
f
)2
. (2.14)
This implies that the trajectory is a circle centered at (x0, y0) and of radius V/|f |. Thesituation is depicted on Figure 2-4.
R = Vf
.....................................................................................................................................
............................................
......................
!V
Ω = f2
(x0, y0)
Figure 2-4 Inertial oscillation of a free
particle on a rotating plane. The or-
bital period is exactly half of the am-
bient revolution period. This figure has
been drawn with a positive Coriolis pa-
rameter, f , representative of the North-ern Hemisphere. If f were negative (asin the Southern Hemisphere), the parti-
cle would veer to the left.
In the absence of rotation (f = 0), this radius is infinite, and the particle follows a straightpath, as we could have anticipated. But, in the presence of rotation (f "= 0), the particle
turns constantly. A quick examination of (2.13) reveals that the particle turns to the right
(clockwise) if f is positive or to the left (counterclockwise) if f is negative. In sum, the ruleis that the particle turns in the sense opposite to that of the ambient rotation.
At this point, we may wonder whether this particle rotation is none other than the negative
of the ambient rotation, in such a way as to keep the particle at rest in the absolute frame of
reference. But, there are at least two reasons why this is not so. The first is that the coordinates
of the center of the particle’s circular path are arbitrary and are therefore not required to
coincide with those of the axis of rotation. The second and most compelling reason is that
the two frequencies of rotation are simply not the same: the ambient rotating plane completes
one revolution in a time equal to Ta = 2π/Ω, whereas the particle covers a full circle in atime equal to Tp = 2π/f = π/Ω, called inertial period. Thus, the particle goes around itsorbit twice as the plane accomplishes a single revolution.
combine and take the square
2.3. FREE MOTION 43
the vector velocity, according to dx/dt = u and dy/dt = v, and a second time integrationprovides
x = x0 −V
fcos(ft + φ) (2.13a)
y = y0 +V
fsin(ft + φ), (2.13b)
where x0 and y0 are additional constants of integration to be determined from the initial
coordinates of the particle. From the last relations, it follows directly that
(x − x0)2 + (y − y0)
2 =
(V
f
)2
. (2.14)
This implies that the trajectory is a circle centered at (x0, y0) and of radius V/|f |. Thesituation is depicted on Figure 2-4.
R = Vf
.....................................................................................................................................
............................................
......................
!V
Ω = f2
(x0, y0)
Figure 2-4 Inertial oscillation of a free
particle on a rotating plane. The or-
bital period is exactly half of the am-
bient revolution period. This figure has
been drawn with a positive Coriolis pa-
rameter, f , representative of the North-ern Hemisphere. If f were negative (asin the Southern Hemisphere), the parti-
cle would veer to the left.
In the absence of rotation (f = 0), this radius is infinite, and the particle follows a straightpath, as we could have anticipated. But, in the presence of rotation (f "= 0), the particle
turns constantly. A quick examination of (2.13) reveals that the particle turns to the right
(clockwise) if f is positive or to the left (counterclockwise) if f is negative. In sum, the ruleis that the particle turns in the sense opposite to that of the ambient rotation.
At this point, we may wonder whether this particle rotation is none other than the negative
of the ambient rotation, in such a way as to keep the particle at rest in the absolute frame of
reference. But, there are at least two reasons why this is not so. The first is that the coordinates
of the center of the particle’s circular path are arbitrary and are therefore not required to
coincide with those of the axis of rotation. The second and most compelling reason is that
the two frequencies of rotation are simply not the same: the ambient rotating plane completes
one revolution in a time equal to Ta = 2π/Ω, whereas the particle covers a full circle in atime equal to Tp = 2π/f = π/Ω, called inertial period. Thus, the particle goes around itsorbit twice as the plane accomplishes a single revolution.
this is the equation of a circle with radius
2.3. FREE MOTION 43
the vector velocity, according to dx/dt = u and dy/dt = v, and a second time integrationprovides
x = x0 −V
fcos(ft + φ) (2.13a)
y = y0 +V
fsin(ft + φ), (2.13b)
where x0 and y0 are additional constants of integration to be determined from the initial
coordinates of the particle. From the last relations, it follows directly that
(x − x0)2 + (y − y0)
2 =
(V
f
)2
. (2.14)
This implies that the trajectory is a circle centered at (x0, y0) and of radius V/|f |. Thesituation is depicted on Figure 2-4.
R = Vf
.....................................................................................................................................
............................................
......................
!V
Ω = f2
(x0, y0)
Figure 2-4 Inertial oscillation of a free
particle on a rotating plane. The or-
bital period is exactly half of the am-
bient revolution period. This figure has
been drawn with a positive Coriolis pa-
rameter, f , representative of the North-ern Hemisphere. If f were negative (asin the Southern Hemisphere), the parti-
cle would veer to the left.
In the absence of rotation (f = 0), this radius is infinite, and the particle follows a straightpath, as we could have anticipated. But, in the presence of rotation (f "= 0), the particle
turns constantly. A quick examination of (2.13) reveals that the particle turns to the right
(clockwise) if f is positive or to the left (counterclockwise) if f is negative. In sum, the ruleis that the particle turns in the sense opposite to that of the ambient rotation.
At this point, we may wonder whether this particle rotation is none other than the negative
of the ambient rotation, in such a way as to keep the particle at rest in the absolute frame of
reference. But, there are at least two reasons why this is not so. The first is that the coordinates
of the center of the particle’s circular path are arbitrary and are therefore not required to
coincide with those of the axis of rotation. The second and most compelling reason is that
the two frequencies of rotation are simply not the same: the ambient rotating plane completes
one revolution in a time equal to Ta = 2π/Ω, whereas the particle covers a full circle in atime equal to Tp = 2π/f = π/Ω, called inertial period. Thus, the particle goes around itsorbit twice as the plane accomplishes a single revolution.
1
2.3. FREE MOTION 43
the vector velocity, according to dx/dt = u and dy/dt = v, and a second time integrationprovides
x = x0 −V
fcos(ft + φ) (2.13a)
y = y0 +V
fsin(ft + φ), (2.13b)
where x0 and y0 are additional constants of integration to be determined from the initial
coordinates of the particle. From the last relations, it follows directly that
(x − x0)2 + (y − y0)
2 =
(V
f
)2
. (2.14)
This implies that the trajectory is a circle centered at (x0, y0) and of radius V/|f |. Thesituation is depicted on Figure 2-4.
R = Vf
.....................................................................................................................................
............................................
......................
!V
Ω = f2
(x0, y0)
Figure 2-4 Inertial oscillation of a free
particle on a rotating plane. The or-
bital period is exactly half of the am-
bient revolution period. This figure has
been drawn with a positive Coriolis pa-
rameter, f , representative of the North-ern Hemisphere. If f were negative (asin the Southern Hemisphere), the parti-
cle would veer to the left.
In the absence of rotation (f = 0), this radius is infinite, and the particle follows a straightpath, as we could have anticipated. But, in the presence of rotation (f "= 0), the particle
turns constantly. A quick examination of (2.13) reveals that the particle turns to the right
(clockwise) if f is positive or to the left (counterclockwise) if f is negative. In sum, the ruleis that the particle turns in the sense opposite to that of the ambient rotation.
At this point, we may wonder whether this particle rotation is none other than the negative
of the ambient rotation, in such a way as to keep the particle at rest in the absolute frame of
reference. But, there are at least two reasons why this is not so. The first is that the coordinates
of the center of the particle’s circular path are arbitrary and are therefore not required to
coincide with those of the axis of rotation. The second and most compelling reason is that
the two frequencies of rotation are simply not the same: the ambient rotating plane completes
one revolution in a time equal to Ta = 2π/Ω, whereas the particle covers a full circle in atime equal to Tp = 2π/f = π/Ω, called inertial period. Thus, the particle goes around itsorbit twice as the plane accomplishes a single revolution.
period of a complete circle is called inertial period
2
Coriolis acceleration in 3D2.5. ROTATING PLANET 49
Ω........................................... ....... ........ ......... .......... .......... .......... .......... .......... .......... ......... ........
.................................................
..........................................................
.................
.................... .....................
...............................
................................. .................................... ..................................... ...................................... ...................................... ...................................... ...................................... .........................................................................
...................................................................................................................................................................................
........... ....... ......... ............... ................... ...................... ......................... ........................... ............................. ............................................. .............................................. .............................................. ............................................. ............................. ........................... ...........................................................................................................
.
....................................
....................................
....................................
....................................
....................................
...................................
..................................
................................
..............................
.....................
...................
..................
..................
...................................
..................................................................
..................
.....................
.........................
............................
..............................
.......
.........................
.......
.......
.......
.......
.......
.......
.......
..
.......
.......
.......
.......
.......
.......
.......
....
.......
.......
.......
.......
.......
.......
.......
....
.......
.......
.......
.......
.......
.......
.......
....
ϕ
λEquator
Parallel
Merid
ian
Greenwich
y
x
z
i
j
k
S
N
Figure 2-9 Definition of a local Carte-
sian framework of reference on a spher-
ical earth. The coordinate x is directedeastward, y northward, and z upward.
du
dt+ 2 Ω × u,
has the following three components
x :du
dt+ 2Ω cosϕ w − 2Ω sin ϕ v (2.20a)
y :dv
dt+ 2Ω sin ϕ u (2.20b)
z :dw
dt− 2Ω cosϕ u. (2.20c)
With x, y, and z everywhere aligned with the local eastward, northward, and vertical direc-tions, the coordinate system is curvilinear, and additional terms arise in the components of
the relative acceleration. These terms will be introduced in Section 3.2, only to be quickly
dismissed because of their relatively small size in most instances.
For convenience, we define the quantities
f = 2Ω sinϕ (2.21)
f∗ = 2Ω cosϕ. (2.22)
2.5. ROTATING PLANET 49
Ω........................................... ....... ........ ......... .......... .......... .......... .......... .......... .......... ......... ........
.................................................
..........................................................
.................
.................... .....................
...............................
................................. .................................... ..................................... ...................................... ...................................... ...................................... ...................................... .........................................................................
...................................................................................................................................................................................
........... ....... ......... ............... ................... ...................... ......................... ........................... ............................. ............................................. .............................................. .............................................. ............................................. ............................. ........................... ...........................................................................................................
.
....................................
....................................
....................................
....................................
....................................
...................................
..................................
................................
..............................
.....................
...................
..................
..................
...................................
..................................................................
..................
.....................
.........................
............................
..............................
.......
.........................
.......
.......
.......
.......
.......
.......
.......
..
.......
.......
.......
.......
.......
.......
.......
....
.......
.......
.......
.......
.......
.......
.......
....
.......
.......
.......
.......
.......
.......
.......
....
ϕ
λEquator
Parallel
Merid
ian
Greenwich
y
x
z
i
j
k
S
N
Figure 2-9 Definition of a local Carte-
sian framework of reference on a spher-
ical earth. The coordinate x is directedeastward, y northward, and z upward.
du
dt+ 2 Ω × u,
has the following three components
x :du
dt+ 2Ω cosϕ w − 2Ω sin ϕ v (2.20a)
y :dv
dt+ 2Ω sin ϕ u (2.20b)
z :dw
dt− 2Ω cosϕ u. (2.20c)
With x, y, and z everywhere aligned with the local eastward, northward, and vertical direc-tions, the coordinate system is curvilinear, and additional terms arise in the components of
the relative acceleration. These terms will be introduced in Section 3.2, only to be quickly
dismissed because of their relatively small size in most instances.
For convenience, we define the quantities
f = 2Ω sinϕ (2.21)
f∗ = 2Ω cosϕ. (2.22)
effective rotation on the sphere
rotation of reference frame
projection on sphere surface
50 CHAPTER 2. THE CORIOLIS FORCE
The coefficient f is called the Coriolis parameter, whereas f∗ has no traditional name andwill be called here the reciprocal Coriolis parameter. In the Northern Hemisphere, f ispositive; it is zero at the equator and negative in the Southern Hemisphere. In contrast, f∗is positive in both hemispheres and vanishes at the poles. An examination of the relative
importance of the various terms (Section 4.3) will reveal that, generally, the f–terms areimportant, whereas the f∗–terms may be neglected.
Horizontal, unforced motions are described by
du
dt− fv = 0 (2.23a)
dv
dt+ fu = 0 (2.23b)
and are still characterized by solution (2.12). The difference resides in the value of f , nowgiven by (2.21). Thus, inertial oscillations on Earth have periodicities equal to 2π/f =π/Ω sinϕ, ranging from 12 h at the poles to infinity along the equator. Pure inertial oscilla-tions are, however, quite rare because of the usual presence of pressure gradients and other
forces. Nonetheless, inertial oscillations are not uncommonly found to contribute to observa-
tions of oceanic currents. An example of such an occurrence, where the inertial oscillations
made up almost the entire signal, was reported by Gustafson and Kullenberg (1936). Cur-
rent measurements in the Baltic Sea showed periodic oscillations about a mean value. When
added to one another to form a so-called progressive vector diagram (Figure 2-10), the cur-
rents distinctly showed a mean drift, on which were superimposed quite regular clockwise
oscillations. The theory of inertial oscillation predicts clockwise rotation in the Northern
Hemisphere with period of 2π/f = π/Ω sin ϕ, or 14 h at the latitude of observations, thusconfirming the interpretation of the observations as inertial oscillations.
2.6 Numerical approach to oscillatory motions
The equations of free motion on a rotating plane (2.11) have been considered in some detail
in Section 2.3, and it is now appropriate to consider their discretization, as the corresponding
terms are part of all numerical models of geophysical flows. Upon introducing the time
increment∆t, an approximation to the components of the velocity will be determined at thediscrete instants tn = n∆t with n = 1, 2, 3, ..., which are denoted un = u(tn) and vn =v(tn), with tildes used to distinguish the discrete solution from the exact one. The so-calledEuler method based on first-order forward differencing yields the simplest discretization of
equations (2.11):
du
dt− fv = 0 −→
un+1 − un
∆t− f vn = 0
dv
dt+ fu = 0 −→
vn+1 − vn
∆t+ fun = 0.
The latter pair can be cast into a recursive form as follows:
equation of inertial oscillations
these describe the unforced motion
Ω =2π
24 hours
Coriolis acceleration in 3D - observations of inertial motions
50 CHAPTER 2. THE CORIOLIS FORCE
The coefficient f is called the Coriolis parameter, whereas f∗ has no traditional name andwill be called here the reciprocal Coriolis parameter. In the Northern Hemisphere, f ispositive; it is zero at the equator and negative in the Southern Hemisphere. In contrast, f∗is positive in both hemispheres and vanishes at the poles. An examination of the relative
importance of the various terms (Section 4.3) will reveal that, generally, the f–terms areimportant, whereas the f∗–terms may be neglected.
Horizontal, unforced motions are described by
du
dt− fv = 0 (2.23a)
dv
dt+ fu = 0 (2.23b)
and are still characterized by solution (2.12). The difference resides in the value of f , nowgiven by (2.21). Thus, inertial oscillations on Earth have periodicities equal to 2π/f =π/Ω sinϕ, ranging from 12 h at the poles to infinity along the equator. Pure inertial oscilla-tions are, however, quite rare because of the usual presence of pressure gradients and other
forces. Nonetheless, inertial oscillations are not uncommonly found to contribute to observa-
tions of oceanic currents. An example of such an occurrence, where the inertial oscillations
made up almost the entire signal, was reported by Gustafson and Kullenberg (1936). Cur-
rent measurements in the Baltic Sea showed periodic oscillations about a mean value. When
added to one another to form a so-called progressive vector diagram (Figure 2-10), the cur-
rents distinctly showed a mean drift, on which were superimposed quite regular clockwise
oscillations. The theory of inertial oscillation predicts clockwise rotation in the Northern
Hemisphere with period of 2π/f = π/Ω sin ϕ, or 14 h at the latitude of observations, thusconfirming the interpretation of the observations as inertial oscillations.
2.6 Numerical approach to oscillatory motions
The equations of free motion on a rotating plane (2.11) have been considered in some detail
in Section 2.3, and it is now appropriate to consider their discretization, as the corresponding
terms are part of all numerical models of geophysical flows. Upon introducing the time
increment∆t, an approximation to the components of the velocity will be determined at thediscrete instants tn = n∆t with n = 1, 2, 3, ..., which are denoted un = u(tn) and vn =v(tn), with tildes used to distinguish the discrete solution from the exact one. The so-calledEuler method based on first-order forward differencing yields the simplest discretization of
equations (2.11):
du
dt− fv = 0 −→
un+1 − un
∆t− f vn = 0
dv
dt+ fu = 0 −→
vn+1 − vn
∆t+ fun = 0.
The latter pair can be cast into a recursive form as follows:
equation of inertial oscillations
these describe the unforced motion
2.6. OSCILLATIONS 51
Figure 2-10 Evidence of inertial os-
cillations in the Baltic Sea, as reported
by Gustafson and Kullenberg (1936).
The plot is a progressive–vector dia-
gram constructed by the successive ad-
dition of velocity measurements at a
fixed location. For weak or uniform
velocities, such a curve approximates
the trajectory that a particle starting at
the point of observation would have
followed during the period of obser-
vation. Numbers indicate days of the
month. Note the persistent veering to
the right, at a period of about 14 hours,
which is the value of 2π/f at that lat-itude (57.8N). [From Gustafson and
Kullenberg, 1936, as adapted by Gill,
1982]
un+1 = un + f∆t vn (2.24a)
vn+1 = vn − f∆t un. (2.24b)
Thus, given initial values u0 and v0 at t0, the solution can be computed easily at time t1
u1 = u0 + f∆t v0 (2.25)
v1 = v0 − f∆t u0. (2.26)
Then, by means of the same algorithm, the solution can be obtained iteratively at times t2, t3
and so on (do not confuse the temporal index with an exponent here and in the following).
Clearly, the main advantage of the preceding scheme is its simplicity, but it is not sufficient
to render it acceptable, as we shall soon learn.
To explore the numerical error generated by the Euler method, we carry out Taylor ex-
pansions of the type
un+1 = un + ∆t
[du
dt
]
t=tn
+∆t2
2
[d2u
dt2
]
t=tn
+ O(∆t3)
Discretizing the intertial oscillation equations
50 CHAPTER 2. THE CORIOLIS FORCE
The coefficient f is called the Coriolis parameter, whereas f∗ has no traditional name andwill be called here the reciprocal Coriolis parameter. In the Northern Hemisphere, f ispositive; it is zero at the equator and negative in the Southern Hemisphere. In contrast, f∗is positive in both hemispheres and vanishes at the poles. An examination of the relative
importance of the various terms (Section 4.3) will reveal that, generally, the f–terms areimportant, whereas the f∗–terms may be neglected.
Horizontal, unforced motions are described by
du
dt− fv = 0 (2.23a)
dv
dt+ fu = 0 (2.23b)
and are still characterized by solution (2.12). The difference resides in the value of f , nowgiven by (2.21). Thus, inertial oscillations on Earth have periodicities equal to 2π/f =π/Ω sinϕ, ranging from 12 h at the poles to infinity along the equator. Pure inertial oscilla-tions are, however, quite rare because of the usual presence of pressure gradients and other
forces. Nonetheless, inertial oscillations are not uncommonly found to contribute to observa-
tions of oceanic currents. An example of such an occurrence, where the inertial oscillations
made up almost the entire signal, was reported by Gustafson and Kullenberg (1936). Cur-
rent measurements in the Baltic Sea showed periodic oscillations about a mean value. When
added to one another to form a so-called progressive vector diagram (Figure 2-10), the cur-
rents distinctly showed a mean drift, on which were superimposed quite regular clockwise
oscillations. The theory of inertial oscillation predicts clockwise rotation in the Northern
Hemisphere with period of 2π/f = π/Ω sin ϕ, or 14 h at the latitude of observations, thusconfirming the interpretation of the observations as inertial oscillations.
2.6 Numerical approach to oscillatory motions
The equations of free motion on a rotating plane (2.11) have been considered in some detail
in Section 2.3, and it is now appropriate to consider their discretization, as the corresponding
terms are part of all numerical models of geophysical flows. Upon introducing the time
increment∆t, an approximation to the components of the velocity will be determined at thediscrete instants tn = n∆t with n = 1, 2, 3, ..., which are denoted un = u(tn) and vn =v(tn), with tildes used to distinguish the discrete solution from the exact one. The so-calledEuler method based on first-order forward differencing yields the simplest discretization of
equations (2.11):
du
dt− fv = 0 −→
un+1 − un
∆t− f vn = 0
dv
dt+ fu = 0 −→
vn+1 − vn
∆t+ fun = 0.
The latter pair can be cast into a recursive form as follows:
Euler Method
52 CHAPTER 2. THE CORIOLIS FORCE
and similarly for v to obtain the following expressions from (2.24)
[du
dt− f v
]
t=tn
= −
[d2u
dt2
]
t=tn
∆t
2+ O(∆t2) (2.27a)
[dv
dt+ fu
]
t=tn
= −
[d2v
dt2
]
t=tn
∆t
2+ O(∆t2). (2.27b)
Derivation of (2.27a) with respect to time and use of (2.27b) to eliminate dv/dt allow us torecast (2.27a) into a simpler form, and similarly for (2.27b):
dun
dt− f vn =
f2∆t
2un + O(∆t2) (2.28a)
dvn
dt+ fun =
f2∆t
2vn + O(∆t2). (2.28b)
Obviously, the numerical scheme mirrors the original equations, except that an additional
term appears in each right-hand side. This additional term takes the form of anti-friction
(friction would have a minus sign instead) and will therefore increase the discrete velocity
over time.
The truncation error of the Euler scheme – the right-hand side of the preceding expres-
sions – tends to zero as ∆t vanishes, which is why the scheme is said to be consistent. Thetruncation is on the order of ∆t at the first power and the scheme is therefore said to befirst-order accurate, which is the lowest possible level of accuracy. Nonetheless, this is not
the chief weakness of the present scheme, since we must expect that the introduction of anti-
friction will create an unphysical acceleratation. Indeed, elementary manipulations of the
time-stepping algorithm (2.24) lead to (un+1)2 + (vn+1)2 = (1+f2∆t2)(un)2 + (vn)2
so that by recursion
‖ u‖2 = (un)2 + (vn)2 = (1 + f2∆t2)n(u0)2 + (v0)2
. (2.29)
So, although the kinetic energy (directly proportional to the squared norm ‖u‖2) of the in-
ertial oscillation must remain constant, as was seen in Section 2.3, the kinetic energy of the
discrete solution increases without bound2 even if the time step ∆t is taken much smallerthan the characteristic time 1/f . Algorithm (2.24) is unstable. Because such a behavior isnot acceptable, we need to formulate an alternative type of discretization.
In our first scheme, the time derivative was taken by going forward from time level tn totn+1 and the other terms at tn, and the scheme became a recursive algorithm to calculate thenext values from the current values. Such a discretization is called an explicit scheme. By
contrast, in an implicit scheme, the terms other than the time derivatives are taken at the new
time tn+1 (which is similar to taking a backward difference for the time derivative):
un+1 − un
∆t− f vn+1 = 0 (2.30a)
vn+1 − vn
∆t+ fun+1 = 0. (2.30b)
2From the context it should be clear that n in (1 + f2∆t2)n is an exponent, whereas in un it is the time index.
In the following text, we will not point out this distinction again, leaving it to the reader to verify the context.
Euler Method Implicit
when rigth-hand side is at future time